Re: Log tools

From: Bradford L. Barrett (brad@mrunix.net)
Date: Mon, May 03 1999


Date: Mon, 3 May 1999 15:09:08 -0400 (EDT)
From: "Bradford L. Barrett" <brad@mrunix.net>
To: www-wca@w3.org
Message-ID: <Pine.LNX.3.96.990503145200.26038G-100000@ns.eurobase.net>
Subject: Re: Log tools


On Mon, 3 May 1999, Johan Hjelm wrote:

> At 14:24 1999-05-03 -0400, Bradford L. Barrett wrote:
> >>
> >I have looked over the proposed metrics you mention, and while some are
> >completely within the scope of what a typical log analysis tool could
> >measure, some are not.  I did not see a great deal of 'recommended
> >methods' on how to measure these metrics either, other than very vague
> >'generic' suggestions on where the data would come from.
> >
> We'd be interested in your input on this, of course. How would you
> structure the list, and what would you include?

I don't have a problem with the list, only it's implementation :)
I realize that conceptually, the metrics mentioned are needed in order to
characterize web growth and usage, however as a developer, I question the
ability to generate some of these metrics in a realistic mannor, based
solely on server logs, and to do so accurately.  For example, under the
"Web Client Metrics", how do I, based on server logs, determine the 
type of client if it reports itself as something I have never seen
before?  Or determine the client-side filtering capability if that data
is not contained in the log?  Of course, you could come up with some
type of cabability matrix based on known clients and what they can/cannot
do, but even that would be inaccurate if for example the client has some
of the built-in capabilities turned off (ie: java, javascript, cookies).

As another example, under "Web Server Metrics", which states the data
comes from log files, I don't see any way to determine, based solely
on logs, things such as Geographical location, number of files/web pages
on the server (you only 'see' the ones that were accessed), etc..
These things simply are not logged currently, so you can't generate
numbers for things that do not exist without some other external imput.

I guess my perspective is from an implementation standpoint, rather than
a theroetical one...  Hey, I have to write code to do this stuff, so I
need to figure out ways to do just that :)

--
Bradford L. Barrett                      brad@mrunix.net
A free electron in a sea of neutrons     DoD#1750 KD4NAW

The only thing Micro$oft has done for society, is make people
believe that computers are inherently unreliable.